Progressive Chess

[hgm]

Does anyone know if an engine for Progressive Chess exists? (This is the variant where white starts with one move, then black plays two moves, white three, blackfour, etc., from an otherwise normal FIDE setup.)

It should be relatively easy to convert a normal Chess engine for this, by tinkering with the turn change. But I have no idea what to use for evaluation (apart from checkmate). Even material doesn't seem very significant. By the time you can make 7 or 8 moves per turn, you might easily win with just a Bishop and a Pawn (by promoting the latter) against a nearly complete army.

My suspicion is that very quickly you get into a regime where it is very difficult to not get checkmated on the next turn. I would like to get an idea of which opening moves are playable.

[Vinvin]

Does anyone know if an engine for Progressive Chess exists? (This is the variant where white starts with one move, then black plays two moves, white three, blackfour, etc., from an otherwise normal FIDE setup.)

It should be relatively easy to convert a normal Chess engine for this, by tinkering with the turn change. But I have no idea what to use for evaluation (apart from checkmate). Even material doesn't seem very significant. By the time you can make 7 or 8 moves per turn, you might easily win with just a Bishop and a Pawn (by promoting the latter) against a nearly complete army.

My suspicion is that very quickly you get into a regime where it is very difficult to not get checkmated on the next turn. I would like to get an idea of which opening moves are playable.

I play around 50 casual games over the board with some friends.
The main idea is : If you can't mate at your turn, you've to prevent be mated after your turn.

[Sven]

Rule questions:
a) What is the definition of checkmate in this game?
b) Do I need to make exactly N moves if it is my turn?

[hgm]

Checkmate is basically the same as in normal chess, because you must solve any checks in the first move of your turn, and are not allowed to pass through check anywhere within your turn.

There are several variants of this game: in Italian Progressive Chess you have to finish all your moves, and delivering check on a non-final move is illegal. (This can lead to a funny 'reverse pinning' of your King, when it is blocking your own slider from capturing the opponent King, and discovering it isan illegalmove!) In Scottish Progressive Chess you can deliver check with any move of the turn, but this then immediately ends the turn (not affecting the number of moves in future turns). So there you usually also save the check for the last move, if you can help it.

[hgm]

I guess the best way to search would be to iteratively deepen the moves in your own turn, to try if you can mate in 1, 2, 3,... And in the mean time score the positions by a conventional King-safety term, counting the number of attacked squares around it, and order the moves in the next iteration accordingly. At the last ply of your turn you would switch to counting material instead (valuing advanced Pawns very high, as they will be Queens for a large part of the opponent turn) and would then traverse the tree within your own turn once more in this order, starting an IID for the full opponent's turn in each leave.

[Vinvin]

In Scottish Progressive Chess you can deliver check with any move of the turn, but this then immediately ends the turn (not affecting the number of moves in future turns). So there you usually also save the check for the last move, if you can help it.

i played only this rule.

[hgm]

This game is apparently played much more seriously than I imagined. I found many hours of videos about it on you tube, discussing opening theory.I had expected it to be a game that is only fun when you play it for the first time without any prior knowledge, but that on deeper analysis it would turn out to be an easy win for one side or the other.

It seems that it is a 'narrow draw', however, according to decades of analysis. I am not sure any computers were used for this analysis, however. The guy who made the videos makes it sound like it was all human effort.

Yet it seems computers would be eminently suited for doing such analysis. Doing a perft(10) for normal Chess is not much of an effort, and it should not be much more difficult if the same side has the move all the time. So it should be easy to check if turns of upto 10 moves can achieve a mate. (Especially since the longer turns would only be played when much of the material has already come off.) You don;t even have to search the whole tree; you can stop at the frst mate you find. And by doing IID based on one-sided King safety, you might be able to find it pretty fast.

[konsolas]

I have a feeling it would all be much simpler than you're implying.

Inside the makemove routine, just only update the side to move when the current side has made the correct number, and only negate beta and alpha when the side to move changes.

[hgm]

Well, yes, that is the basic idea. But that still leaves very important details. Like how to deepen the search. Would you search 1, 3, 6, 10, 15, 21, ... ply, to make each side finish its turn? Or would you search 1, 2, 3, 4, 5, ... ply, ending turns half-way? In the latter case, how would you evaluate half-way during a turn? The number of moves you still have to go before your turn is done seems one of the biggest assets of a player, but its exact value in relation to material seems very much dependent on how powerful your own material is, and how much material there is to grab for you.

The point is that the two major goals of the game need some what opposite means to achieve. You want to checkmate when you can, and don't care at all about the material balance in that case (or your own King safety). To efficiently find a mate in a many-move sequence it would enormously help to search moves that compromise the opponent's King safety first (attacking squares neighboring the King). If you fail to deliver mate, your primary concern becomes not getting mated on the next turn. The best move for compromising the opponent's King safety usually do nothing to improve your own, and thus suddenly become very poor moves if in the end they fall short of achieveing the checkmate goal. Sofor the purposeof surviving you would like to search the moves of your own turn in a completely different order. One of the most important factors in asuring your King's survival is to remove as much of the opponent material that threatens it as you can. In any case Queens and Pawns close to promotion.

So it seems some Internal Iterative Deepening is in order, where at the start of a (say) 5-ply turn you would first search 1, 2, 3, 4 ply with an evaluation only scoring opponent King safety, and then switch to an evaluation aimed at removing material and improving own King safety otherwise for the 5-ply search, so that the tree will be traversed in a completely different order when you do the 11-ply search next (where deepening will be controlled in the node at the start of the 6-ply turn).

[Gerd Isenberg]

I guess you are already aware of
Vito Janko, Matej Guid (2015). Development of a Program for Playing Progressive Chess. Advances in Computer Games 14
and their program
https://ailab.si/progressive-chess/